Abstract: We apply borehole strainmeter data with 10^-10～10^-8 resolution to analyse strain changes before the 2004 Macquarie Island M_S8.0 and 2004 Sumatra M_S8.9 earthquakes, and to quantitatively and objectively test precursory waves using high-quality observations at 13 stations. First, to remove solid tide with period longer than 128 minutes, we filtered data with high-pass filter. Secondly, we remove air-pressure influence by linear regression method. Finally, we apply wavelet transform to analyze the high-pass filtered data, decomposing the signal within the period of 2～128 minutes into 6 different frequency bands. The result indicates that wavelet transform can be used to identify small changes existed in the deformation signal. To quantitatively analyze the characteristic precursory waves, we proposed an overrunning points ratio method to calculate overrun points in certain times. We analysed all the 6 ferequency bands and calculated the slopes of regression line, the averages, the standard deviation and the ratios of the slope that are greater than 0 and less than 0. These three kinds of data show that precursory waves can not be generally detected in the period of 15 days before the two strong earthquakes.

Key words: Borehole strainmeter observations, Precursory wave, Wavelet transform, Overrun points ratio analysis